SanDisk: 3D flash chips? Sure, buddy, we'll see you in 2016
The tech just doesn't stack up... for now
Electronics giant SanDisk will be deliberately late in building 3D flash dies. It believes current NAND chip technology has two more generations left before going 3D becomes cost-effective.
Like every other NAND fab operator, SanDisk's strategy is to shrink the current 2D NAND dies as small as possible and then start stacking dies on top of each other to create 3D NAND, and so carry on growing the number of flash bits you can get in a two-dimensional footprint.
The roadmap was revealed in SanDisk's Q2 2013 results earnings call by CEO Sanjay Mehrotra.
Our view of SanDisk's summarised NAND technology roadmap is this:
- 19nm in production now
- 1Y in Q3 2013
- 1Z in Q4 2014/Q1 2015
- BiCS 3D NAND - Q4 2015/Q1 2016
Shrinking the cell size means more dies will fit on a standard wafer, from which individual dies are cut; the cost per die thus decreases, as wafer costs are fairly static. A second consequence is that amount of NAND flash you get in a standard-sized chip increases as you can cram more cells into the chip.
The terminology is that 19nm is known as 1X technology, located on a scale from 19nm to 10nm; 1Y is some value between 19nm and 15nm, and 1Z is somewhere below that. The values are relative; 1Z being smaller than 1Y, and 1Y being smaller than 1X.
Each flash fab operator's definition of 1X, 1Y and 1Z may vary, except that 1X can be no larger than 19nm and 1Z can be no smaller than 10nm.
As the cell size shrinks, the performance and reliability of the NAND reduces such that it is generally agreed that NAND with cell sizes less than 10nm will be uneconomic to make. Yet demand for increased NAND capacity will continue, and the way to meet that demand is to use the same fundamental NAND technology but stacking dies on on top of the other so as to, literally, build up capacity.
Micron is sampling 16nm NAND now. One view is that it will not produce a smaller-than-16nm cell and will move straight to 3D NAND. SanDisk believes, perhaps differing from other flash foundry folks, that they can have three bites of the 19nm-to-10nm flash cherry, not just one or two, and then go 3D using its Bit Cost Scalable (BICS) technology, invented by its partner Toshiba some years ago.
The costs of making 3D NAND will be higher than 2D NAND because, for one thing, checking individual cell operations when they are buried under other cells will be hard. SanDisk is calculating that if it can make its 1Y and 1Z NAND technology work then it will be more profitable than costlier 3D NAND, allowing it to offer equivalent or near-equivalent capacity to 3D NAND in the same footprint.
Oh, and SanDisk's quarterly results? They were good. Revenues stood at $1.48bn, up 43 per cent year on year and 10 per cent higher than the previous quarter. Profits came in at $262m, massively higher than the $13m recorded a year ago and greatly higher than the previous quarter's $166 million.
SanDosh SanDisk thinks it's on track for record revenues this year. ®